机构-信息-公司投资Agency, Information and Corporate Investment

Agency, Information and Corporate Investment
Jeremy C. Stein
Harvard University and NBER
June 2001
Abstract: This essay surveys the body of research that asks how the efficiency of corporate investment is influenced by problems of asymmetric information and agency. I organize the material around two basic questions. First, does the external capital market channel the right amount of money to each firm? That is, does the market get across-firm allocations right, so that the marginal return to investment in firm i is the same as the marginal return to investment in firm j? Second, do internal capital markets channel the right amount of money to individual projects within firms? That is, does the internal capital budgeting process get within-firm allocations right, so that the marginal return to investment in firm i’s division A is the same as the marginal return to investment in firm i’s division B? In addition to discussing the theoretical and empirical work that bears most directly on these questions, the essay also briefly sketches some of the implications of this work for broader issues in both macroeconomics and the theory of the firm.
This paper is to appear as a chapter in the Handbook of the Economics of Finance, edited by George Constantinides, Milt Harris and René Stulz. I am grateful to the NSF for financial support, and to Geoff Tate and Ann Richards for research assistance. Thanks also to Judy Chevalier, Oliver Hart, Bengt Holmstrom, Steve Kaplan, Owen Lamont, Raghu Rajan, David Scharfstein, Andrei Shleifer and René Stulz for their input.
I. Introduction
A fundamental question in corporate finance is this: to what extent does capital get allocated to the right investment projects? In a perfect world, with frictionless capital markets of the sort envisioned in Modigliani and Miller (1958), funds flow in such a way that the marginal product of capital is equated across every project in the economy. Of course, in the real world, there are a variety of distortionary forces that prevent things from working this well. Taxes and transactions costs are examples of such frictions. But perhaps the most pervasive and important factors influencing the efficiency of corporate investment are those that arise from informational asymmetries and agency problems.
This essay surveys research–both theoretical and empirical–that speaks to the influence of asymmetric information and agency on investment behavior. I organize the material by noting that the fundamental question posed above can be divided into two sub-questions. First, does the external capital market channel the right amount of money to each firm? In other words, does the market get across-firm allocations right, so that the marginal return to investment in firm i is the same as the marginal return to investment in firm j?
Second, do internal capital markets channel the right amount of money to individual projects within firms? In other words, does the internal capital budgeting process get within-firm allocations right, so that the marginal return to investment in, say, firm i’s division A is the same as the marginal return to investment in firm i’s division B?
Although these two questions are logically distinct–in the sense that the workings of the external capital market appear in many ways to be quite different from those of the internal
1
capital market–an overarching goal of this essay is to emphasize the common elements of the capital-allocation problem across and within firms. For example, just as investors in the external capital market have to be wary of dealing with a CEO who is better informed about firm prospects than they, and whose incentives diverge from theirs, so must a CEO overseeing the internal capital budgeting process be wary of dealing with subordinates who are better informed about divisional prospects than she, and whose incentives diverge from hers. While the external capital market may ultimately resolve this problem through different means than the internal capital market–with different consequences for investment behavior–it is nevertheless important to appreciate that the underlying problem may well be the same one in both cases.
Both of the sub-questions have been the subject of extensive theoretical and empirical work. Still, it is fair to say that research on the first sub-question–that having to do with the efficiency of across-firm capital allocation–is currently at a more mature stage. On the notion that life is more exciting near the frontier, I will thus devote a somewhat disproportionate share of my attention to surveying work on the second sub-question, that of within-firm capital allocation. On the first, and especially when it comes to empirical work, I will defer more to existing survey papers (e.g., Hubbard (1998)).
Scope of the essay: what’s covered and what’s left out
As much as possible, I am going to focus on research that speaks directly to the impact of information and agency problems on investment behavior. To oversimplify, but not by much, most of the empirical papers that I will touch on have some measure of investment as the left-hand-side variable. Of course, the concepts of asymmetric information and agency are central to
2
virtually every major topic in corporate finance, including corporate governance, capital structure, the design of incentive contracts, financial intermediation, etc. Indeed, one can think of governance, capital structure, incentive contracts and intermediation as a variety of curative mechanisms that arise endogenously to mitigate the effects of information and agency problems on investment outcomes. Thus, at some level, it is difficult to satisfactorily address the subject of investment without taking on these other topics as well.
Nevertheless, although this will no doubt lead to some awkwardness and many omissions, I will for the most part leave these curative mechanisms lurking in the background.1 This can be thought of as a partial equilibrium approach, where it is implicitly assumed that certain types of information and agency distortions are not fully resolved by the curative mechanisms, and
thus–for reasons that are exogenous to the model–remain relevant in equilibrium. This partial equilibrium approach is the only way I can think of to keep the scope of this essay manageable.
Moreover, in much of what follows, I will give primary emphasis to those types of investment distortions that are the most pervasive and stubborn, in the sense that they are likely to exist even when agency and information problems are relatively “mild”–that is, even when the legal, auditing, and contracting environment is highly evolved. (Think of the U.S. environment, for example). I will have less to say about more extreme distortions that arise in economies and situations where investors are poorly protected, and managers can loot, tunnel, etc.2 1Fortunately, there are already several surveys on these topics. In addition to the essays in this volume, see, e.g., Shleifer and Vishny (1997) on governance, and Harris and Raviv (1991) on capital structure.
2See Johnson, LaPorta, Lopez-de-Silanes and Shleifer (2000) for several examples of tunneling. Of course, even in economies such as the U.S. where it is not often observed in equilibrium, the out-of-equilibrium threat of such very bad behavior may do a lot to explain
3
Finally, although I will discuss the general consequences of high leverage for investment, I will not address the details of how financially distressed companies restructure their assets, either inside or outside of formal bankruptcy. So perhaps the best way to interpret much of what I am doing is to think of a financially healthy firm operating in an environment where governance and other curative mechanisms are about as good as they can be, and to ask: what can still go wrong?
Organization
The remainder of this essay is divided into two main parts. Part One deals with investment at the firm level, and contains three sections. I begin in Section II by reviewing the various major classes of theories that are relevant for understanding investment at the firm level. In Section III, I discuss the empirical evidence that speaks to these theories. In Section IV, I touch briefly on the macroeconomic implications of this research.
Part Two of the essay deals with investment inside firms. Section V covers the theoretical work, and Section VI the associated empirical work.
Finally, in Section VII, I conclude by offering some tentative thoughts on how the central ideas in the essay can be used to think about the boundaries of the firm.
various features of governance, law, disclosure policies, etc.
4
PART ONE: INVESTMENT AT THE FIRM LEVEL
II. Theoretical Building Blocks: Investment at the Firm Level
There are many, many theoretical models that have implications for investment at the firm level, and there a variety of ways that one could go about grouping them. For the purposes of the discussion that follows, I will take an empirically-oriented approach to organizing the theories. That is, I will cluster together those models that have similar empirical implications, even if the underlying theoretical mechanisms are quite distinct. The converse and potentially awkward feature of this approach is that sometimes models that are quite close in terms of their underlying logic will get placed into different categories. To take a concrete example, the models of Myers (1977) and Hart and Moore (1995) are both built on the same foundation–the idea that a large debt burden can prevent a company from raising the funds to undertake new investment. But in the former paper, managers are benevolent towards outside shareholders, and there is always underinvestment in equilibrium; in contrast, in the latter, managers are self-interested and there can be either underinvestment or overinvestment, depending on the state of the world. Thus, although the formal structure of these models is quite similar, I will put them into different groupings.
II.A. Models of Costly External Finance
The first broad class of models to be considered are those that unambiguously predict underinvestment relative to a first-best benchmark. In these models, managers can for the most
5
part be thought of as acting in the interests of current shareholders, at least in equilibrium.3 Thus when managers have access to unlimited discretionary resources, investment converges to the efficient level. However, when managers are resource-constrained in some way or another, there will be too little investment, because there are frictions associated with raising finance externally.
II.A.1. Costs of equity finance
An important insight, due to Myers and Majluf (1984), Myers (1984) and Greenwald, Stiglitz and Weiss (1984), is that raising equity externally will generally be problematic due to an adverse-selection problem of the sort first identified by Akerlof (1970).4 To the extent that managers favor their current stockholders at the expense of potential future investors, they will wish to sell new shares at times when their private information suggests that these new shares are most overvalued. As a result, equity issues are rationally interpreted by the market as bad news (Asquith and Mullins (1986), Masulis and Korwar (1986), Mikkelson and Partch (1986)), which in turn can make managers of good firms (those with high realizations of their private information) reluctant to sell equity in the first place. The bottom line is that even firms who are badly in need of new equity–say because they have good investment opportunities but scarce internal resources–may be unable or unwilling to raise it.
3Though in some cases, (e.g., Townsend (1979), Gale and Hellwig (1985), Bolton and Scharfstein (1990), Hart and Moore (1998)) managers act on behalf of shareholders only because they are in equilibrium the only shareholders. In these entrepreneurial-firm models, agency problems are so severe as to rule out the use of outside equity finance.
4The Myers-Majluf model has been extended and refined by many authors, (e.g., Krasker (1986)). See Harris and Raviv (1991) for a discussion and references. Dybvig and Zender (1991) have questioned the microfoundations of the assumption that managers act on behalf of existing shareholders, while Persons (1994) has offered a rationalization of this assumption.
6
II.A.2. Costs of debt finance
Of course, an inability to access new equity would not compromise investment if firms could frictionlessly raise unlimited amounts of debt financing. However, a variety of theories suggest that this is unlikely to be the case.
II.A.1.a. adverse selection, moral hazard and credit rationing in the debt market
The same basic adverse selection argument that is used by Myers and Majluf (1984) for the equity market can be applied to the debt market, to the extent that the debt involved has some default risk: at any given interest rate, managers will be more likely to borrow if their private information suggests that they are relatively prone to default. Or, as a variation on the theme, there can be moral hazard, whereby those managers who borrow have an increased incentive to take the sort of risks that lead to default. As has been shown by Jaffee and Russell (1976), Stiglitz and Weiss (1981, 1983) and others, these sorts of considerations can lead to credit rationing, whereby firms are simply unable to obtain all the debt financing they would like at the prevailing market interest rate.5
II.A.2.b. debt overhang
Myers (1977) is another paper that speaks to the limitations of debt finance. Here the problem is not so much in accessing the debt market ex ante, but rather in what happens after the money is borrowed. In particular, a large debt burden on a firm’s balance sheet discourages 5In spite of the similarities, Myers (1984) and Myers and Majluf (1984) argue that adverse selection problems are generally likely to be more severe in the equity market, because equity values are more sensitive than debt values to managers’ private information.
7
further new investment, particularly if this new investment is financed by issuing claims that are junior to the existing debt. This is because if the existing debt is trading at less than face value, it acts as a tax on the proceeds of the new investment: part of any increase in value generated by the new investment goes to make the existing lenders whole, and is therefore is unavailable to repay those claimaints who put up the new money.6
Debt overhang models can be thought of as having two distinct sorts of empirical implications: ex post (once the debt burden is in place) they suggest that highly-leveraged firms, such as those that have recently undergone leveraged buyouts, will be particularly prone to underinvestment. Ex ante, they offer a reason why even more modestly-levered firms, particularly those with attractive future investment opportunities, may be reluctant to raise much debt in the first place, even if this means foregoing some current investment projects.7
II.A.2.c. optimal contracting models of debt: underinvestment in entrepreneurial firms
The above-discussed models of debt and equity finance take the existence of these types of financial claims as given, and then go on to derive implications for investment, capital structure, etc. Another branch of the literature seeks to endogenize the financial contract,
6The basic debt overhang concept has proved to be enormously useful in addressing a wide range of questions having to do with: i) debt structure (seniority, security, etc.); as well as ii) the more specific details of how financial distress plays itself out and is resolved. For a few examples from a very large literature, see Stulz and Johnson (1985), Berkovitch and Kim (1990), Bergman and Callen (1991) Hart and Moore (1995), and Gertner and Scharfstein (1991). Again, see Harris and Raviv (1991) for more complete references.
7Jensen and Meckling (1976) offer another reason why firms might be unwilling to take on too much debt ex ante: an excessive debt burden can create incentives for managers, acting on behalf of shareholders, to take on risky negative-NPV projects at the expense of lenders.
8
typically by positing some specific agency problem (e.g., managers’ penchant for diverting the firm’s cashflow to themselves) and asking what sort of claim represents an optimal response to this agency problem.
In much of this work, the optimal contract that emerges resembles a standard debt contract, and there is no outside equity financing.8 Thus the firms in question should be interpreted as “entrepreneurial”, in the sense that their only stockholders are their managers. Early examples include Townsend (1979) and Gale and Hellwig (1985), who assume that outside investors can only verify a firm’s cashflows by paying some fixed auditing cost. As long as the manager turns over the stipulated debt payments, there is no audit, and the manager gets to keep the rest of the firm’s cashflow. However, if the manager fails to make the debt payment, the lender audits, and keeps everything he finds; this can be interpreted as costly bankruptcy. The implications for investment follow from the auditing/bankruptcy cost. In particular, the less wealth the manager is able to put up, and hence the more he must borrow, the greater is the likelihood of the auditing cost being incurred. Thus less managerial wealth translates into greater deadweight costs of external finance, and less investment.
More recently, following the work of Grossman and Hart (1986), Hart and Moore (1990), and Hart (1995) on incomplete contracting, the emphasis has shifted to thinking of financial contracts in terms of the allocation of control rights that they embody; Aghion and
8Debt tends to be an attractive contract when verification of cashflows is costly or impossible, so that managers have broad scope for diverting these cashflows to themselves. However, Fluck (1998) and Myers (2000) show how outside equity financing can also be sustained in such a setting, provided there is an infinite horizon. (See also Gomes (2000) for a related argument.) In other cases, when cashflows can be more readily verified, optimal financing schemes can involve a richer mix of claims. See, e.g., Dewatripont and Tirole (1994).
9
Bolton (1992) were among the first to take this point of view. In this context, debt is often seen as an incentive scheme that rewards management with continued control if it makes the required debt payments, and punishes it with loss of control otherwise. In a multi-period framework, this type of incentive scheme enables outside lenders to extract payments from managers even in the extreme case where cashflows are completely unverifiable. Well-known papers in this vein include Bolton and Scharfstein (1990), and Hart and Moore (1994, 1998).
Like the costly-state-verification models, these models also have the feature that there is underinvestment ex ante, with this underinvestment problem being a decreasing function of managers’ wealth. Moreover, given the multi-period nature of the models, one can also interpret some of them as implying a form of ex-post underinvestment as well, with assets sometimes being prematurely seized and liquidated by lenders when managers are unable to meet their debt payments.9
II.A.3. Synthesis: A Reduced-Form Model of Costly External Finance
In spite of the wide variety of modeling approaches, all the theories surveyed thus far have broadly similar empirical implications for investment. Indeed, the essence of what these theories have to say about investment can be captured in a very simple reduced-form model. Although the model may appear ad hoc, Froot, Scharfstein and Stein (1993) demonstrate that it can be mapped precisely into a variant of the Townsend (1979) and Gale-Hellwig (1985) costly-state-verification models. Also, Stein (1998) shows that an appropriately parameterized version of the Myers-Majulf (1984) adverse-selection model leads to essentially the same reduced form.
9See also Diamond (1991) for a model with excessive ex post liquidation by lenders.
10
The setup is as follows. The firm invests I at time 1, which yields a gross return of f(I) at time 2, where f( ) is an increasing, concave function. Of the investment I, an amount w is financed out of internal resources (managers’ wealth, or the firm’s retained earnings) and an amount e is raised externally, via new issues of debt, equity or some other claim. Thus the budget constraint is I = e + w. In a first-best world, managers would seek to maximize:
max f(I)/(1 + r) - I(1)
where r is the risk-adjusted discount rate. This involves setting the marginal product of capital, , equal to (1 + r).
f
I
One can loosely capture some of the financing frictions discussed above by assuming that there are deadweight costs associated with funds raised externally, and that these costs are given by C(e), where C( ) is an increasing convex function, and is a measure of the degree of the financing friction. Thus the firm’s problem becomes:
max f(I)/(1 + r) - I - C(e)(2)
Kaplan and Zingales (1997) show that the solution in this case has the following properties. First, I is always less than or equal to the first best. Also, dI/dw 0 and dI/d 0: I is (weakly) increasing in the firm’s internal resources w, and (weakly) decreasing in the degree of the financing friction . These features are exactly what one would expect. However, there is more subtlety in the behavior of some the higher-order derivatives of I. In particular, d2I/dw2
11
cannot be unambiguously signed. Thus while the local sensitivity of investment to internal cash, dI/dw, eventually converges to zero for w high enough, this convergence need not be monotonic. Similarly, one cannot in general sign d2I/dwd . As Kaplan and Zingales (1997) emphasize, the important message for empirical work is that one has to be careful in using measures of dI/dw as proxies for . That is, in comparing two firms, it is not necessarily true that the one with the higher empirically-measured sensitivity of investment to internal cash should be thought of as the one facing the more severe financing frictions. I will return to this caveat below.
II.B. The Agency Conflict Between Managers and Outside Stockholders
In the models discussed so far, there is in equilibrium no meaningful conflict between managers and stockholders. This is either because managers are simply assumed to act in the interests of stockholders (as in Myers and Majluf (1984), and Myers (1977)) or, at the other extreme, because the threat of managerial expropriation of outside stockholders is so great that equity financing is not viable in equilibrium, and the firm remains owner-managed. But a central theme in much of the corporate-finance literature–with a lineage going back to Berle and Means (1932), and including the influential work of Jensen and Meckling (1976)–is that the managers of publicly-traded firms pursue their own private objectives, which need not coincide with those of outside stockholders.
There are many possible manifestations of the manager-stockholder agency conflict. For example, managers may simply not exert as much effort as they would in a first-best world (Holmstrom (1979)). Given the focus of this essay, however, I restrict attention to those variants of the agency problem that have the most direct implications for investment.
12
II.B.1. Empire-building
II.B.1.a. Empire-building and overinvestment
One way in which managers’ interests may diverge from those of stockholders is that managers may have an excessive taste for running large firms, as opposed to simply profitable ones. This “empire-building” tendency is emphasized by Baumol (1959), Marris (1964), Williamson (1964), Donaldson (1984) and Jensen (1986, 1993), among many others.
Jensen (1986, 1993) argues that empire-building preferences will cause managers to spend essentially all available funds on investment projects. This leads to the prediction that investment will be increasing in internal resources. It also implies that investment will decrease with leverage, because high current debt payments force cash out of the firm, thereby reducing managers’ discretionary budgets. Note that these are the same basic predictions that emerge from the costly-external-finance genre of models described in Section II.A above, though of course the welfare implications are very different.
Jensen’s ideas have been further developed and refined in formal models by Stulz (1990), Harris and Raviv (1990), Hart and Moore (1995), and Zwiebel (1996).10 These models typically incorporate empire-building preferences by using the modeling device of managerial private benefits of control (Grossman and Hart (1988)), and assuming that these private benefits are proportional to either the amount the firm invests (Hart and Moore (1995)), or the gross output
10With respect to the general idea that debt can serve as a disciplinary device, an important precursor to these papers is Grossman and Hart (1982).
13
from investment (Stulz (1990)).11 One insight that comes from these models is that no matter how strong the underlying agency problem, it would be wrong to conclude that empire-building tendencies necessarily lead to an empirical prediction of overinvestment on average. Rather, the usual outcome in the models is an endogenously determined level of debt that attempts to balance ex post over- and underinvestment distortions. Thus the models predict ex post overinvestment in some states of the world (when the level of free cashflow relative to investment opportunities is higher than expected), and ex post underinvestment in others.
As a very loose heuristic way of comparing the empirical content of empire-building models to those of costly external finance, one can modify equation (2) above in the spirit of Stulz (1990) and Hart and Moore (1995) by adding a term equal to f(I) to the objective function. This captures the idea that managers derive private benefits from gross investment output, as in Stulz (1990), with measuring the intensity of the agency conflict. Thus (2) becomes:
max (1 + )f(I)/(1 + r) - I - C(e)(3)
As internal resources w go to infinity, the marginal product of capital now asymptotes at (1 + r)/(1 + ), rather than at (1 + r)–i.e., there is overinvestment. However, more generally, there can be either over- or underinvestment, depending on the realization of w relative to other
11The latter formulation–private benefits proportional to output–implies that managers overinvest, but that conditional on the level of investment, they rank projects in the right order, from high to low NPV. This seems to capture the behavior described by Donaldson (1984).
14
parameter values.12 And importantly, most of the other comparative statics of the model–having to do with dI/dw, dI/d , d2I/dw2 and d2I/dwd –are the same as before. Again, this underscores the challenges associated with empirically distinguishing the two classes of theories.13
II.B.1.b. Empire-preservation, entrenchment and diversification
If managers do in fact derive private benefits from being in charge of large corporate empires, this is likely to show up not just as an overall tendency toward overinvestment. Rather, some specific types of investments will seem especially attractive to managers. For example, Amihud and Lev (1981) argue that there will be a managerial preference for diversification, as this reduces the risk of the empire going out of business. And Shleifer and Vishny (1989) suggest that managers will be particularly keen to invest in projects that require their specific human capital, thereby strengthening their chances of keeping their jobs.
II.B.2. Reputational and career concerns
Another source of conflict between managers and shareholders is that managers may be concerned with how their actions affect their reputations, and ultimately their perceived value in
12The models discussed above suggest that w will be in part endogenously determined by the firm’s choice of capital structure policy.
13Hadlock (1998) argues that empire-building models have the property that dI/dw is decreasing in managerial incentives, while a costly-external-finance model of the Myers-Majluf (1984) type has the opposite property, and uses this insight to construct a differentiating test.
15。